2

Is there a way to guarantee redundancy to an SSD using an HDD?

Raid 1 is the best choice to guarantee redundancy in HDDs, but SSDs are too expensive to guarantee redundancy via RAID. If I was to couple an SSD with an HDD, could I guarantee redundancy using the HDD as a failover device, and lazily mirroring the data on the HDD? (e.g.: every 5 minutes the data should be synchronized, rather than in real time like with Raid 1).

Mascarpone
  • 872
  • 3
  • 9
  • 27
  • Failure from I/O is a lot lower on SSD disks. If this is for backup purposes use a backup solution instead of thinking of it as redundancy. – Matthew Ife Nov 12 '11 at 21:36

2 Answers2

5

If every 5 minutes is enough and there are not too many writes rsync is your fried here.

Nils
  • 7,657
  • 3
  • 31
  • 71
4

The use-case here is flawed. The one use-case where this makes sense is where your I/O activity is majority-read by a very large degree (like 9:1). In that case, since there are so few writes the HD can keep up with the changes coming from the SSD. If the ratio is any less then that, then the SSD risks outpacing the HD's entire write ability.

What you're looking at isn't RAID, it's replication. Something like is what you're looking for. The system will keep a delta-cache of writes to commit to the HD, and periodically flush those writes to the HD. But as I said, since an SSH can handle w-a-y more writes than an HD can, this only works if you're doing very little writing to the SSD.

sysadmin1138
  • 131,083
  • 18
  • 173
  • 296
  • drdb is often used to get RAID1-like mirroring behavior, I would put it closer in the block-level mirroring camp than replication. – kbyrd Nov 12 '11 at 18:14
  • Is there any example for using DRBD as local async raid1? It does some strict tests on the IP-level during startup... – Nils Nov 12 '11 at 21:29